JingweiJ / ActionGenome

A video database bridging human actions and human-object relationships
MIT License
127 stars 17 forks source link

Question about the annotations #2

Open coldmanck opened 4 years ago

coldmanck commented 4 years ago

Thanks again for the wonderful work. Regarding the persons & objects annotation, I appreciate if you could clarify my questions below:

image

JingweiJ commented 4 years ago

Hi @coldmanck

  1. Yes for the current version with person detection only (not really annotation). We're annotating person boxes for multiple persons.
  2. Yes. We've asked the annotators to only localize the object that is being interacted at each moment. E.g. for action "sitting on a chair", only the chair that is being interacted will be localized, rather than all chairs in the scene if there are multiple of them.
  3. This is also a problem when multiple persons are representing in the scene and doing different actions. As we're annotating person boxes, hopefully this issue will be addressed in the next release.
coldmanck commented 4 years ago

Hi @JingweiJ

Thank you for the response. Just wondering, do you have a rough time schedule for the next release? I am asking as I would like to make use of the dataset but the current annotations don’t seem to work for me.